Flying foxes optimization with reinforcement learning for vehicle detection in UAV imagery

Area of Science:

• Computer Science
• Artificial Intelligence
• Transportation Engineering

Background:

• Intelligent Transportation Systems (ITS) are crucial for smart cities, integrating autonomous and connected vehicles.
• Unmanned Aerial Vehicles (UAVs) are increasingly used in ITS for surveillance and data collection.
• Accurate detection and classification of on-ground vehicles from aerial imagery are vital for traffic management, disaster response, and navigation.

Purpose of the Study:

• To develop an automated system for detecting and classifying vehicles in aerial images for ITS applications.
• To enhance the accuracy and efficiency of vehicle identification in complex urban and disaster-affected environments.
• To leverage deep learning and optimization techniques for robust vehicle analysis from UAV-captured data.

Main Methods:

• The proposed Flying Foxes Optimization with Deep Learning-based Vehicle Detection and Classification on Aerial Images (FFODL-VDCAI) technique.
• Vehicle detection using YOLO-GD, incorporating Ghost-Net and Depthwise convolution for efficiency.
• Hyperparameter tuning of Ghost Net via the Flying Foxes Optimization (FFO) algorithm.
• Vehicle classification using a Deep Q-Network (DQN) based reinforcement learning approach.

Main Results:

• The FFODL-VDCAI methodology demonstrated superior performance in vehicle detection and classification tasks.
• Achieved high accuracy rates of 96.15% on the PSU dataset and 92.03% on the Stanford dataset.
• Validated the effectiveness of the integrated deep learning and optimization approach on UAV image datasets.

Conclusions:

• The FFODL-VDCAI technique provides an effective and accurate solution for automated vehicle detection and classification in ITS.
• The integration of YOLO-GD, FFO, and DQN offers a powerful framework for analyzing aerial imagery from UAVs.
• This research contributes to advancing smart city infrastructure by improving real-time transportation monitoring capabilities.